最优子集的神经网络预报建模研究

作者尝试用最优子集方法进行神经网络长期预报模型的建模方法研究。结果表明，在很多情况下，由于最优子集方法比逐步回归方法能选取更好的预报因子，因此所构造的神经网络预报模型具有更好的拟合和预报效果，这为神经网络在长期预报的应用研究提供了新的思路和方法。

Study of the Prediction Modeling of Neural Network Using the Optimal Subset

A method about the neural network long tem forecast is studied by using the optimal subset. The results show that the neural network model established by the optimal subset is found of better fitting and prediction accuracy, because the method using the optimal subset can select out better factors than stepwise regression. Thus it provides a new line for the research of neural network prediction on long term forecast.